資源描述:
《帶變異算子pso算法在酶發(fā)酵控制中的優(yōu)化研究》由會員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、摘要課題的依托背景是“十五”后期國家高技術(shù)研究發(fā)展計(jì)劃(863)基金資助項(xiàng)目?功能基因亞功能片段的制各和質(zhì)量控制技術(shù)平臺。為了保證整個(gè)系統(tǒng)的高質(zhì)高產(chǎn),必須優(yōu)化發(fā)酵過程,提高發(fā)酵的產(chǎn)量和質(zhì)量,所以課題重點(diǎn)放在帶變異算子Ps0算法在酶發(fā)酵的補(bǔ)料速率只、pⅣ值和Dp控制中的優(yōu)化研究上。粒子群優(yōu)化算法(PanicleswarmOptimization,PsO)算法與其它算法相比,具有更快的收斂速度,操作更為簡單。本文針對標(biāo)準(zhǔn)Ps0算法處理高維、復(fù)雜問題時(shí),易于陷入早熟收斂的缺陷進(jìn)行改進(jìn),提出帶變異算子的粒子群優(yōu)
2、化算法VPSO(Ⅷiation.OperatorPS0),經(jīng)過多個(gè)測試函數(shù)的測試證明,VPSO算法具有較快的全局收斂速度和強(qiáng)大的全局搜索能力,能有效的避免早熟收斂問題,尋優(yōu)效果明顯好于標(biāo)準(zhǔn)Ps0算法和遺傳算法。酶優(yōu)化控制方案的設(shè)計(jì)是基于BP神經(jīng)網(wǎng)絡(luò)的預(yù)測控制和VPSO算法的參數(shù)尋優(yōu),鑒于標(biāo)準(zhǔn)BP算法收斂太慢的缺點(diǎn),運(yùn)用PsO算法來優(yōu)化網(wǎng)絡(luò)權(quán)值,通過仿真實(shí)驗(yàn)表明,此算法無論是收斂速度,還是避免早熟收斂,均取得不錯的效果。為了驗(yàn)證此優(yōu)化控制方案的可行性與有效性,將其實(shí)際應(yīng)用于對L天冬酰胺酶II發(fā)酵的過程控制
3、,實(shí)踐結(jié)果表明,優(yōu)化控制方案的設(shè)計(jì)比較成功,與優(yōu)化控制前的那些高產(chǎn)發(fā)酵結(jié)果相比,平均縮短發(fā)酵時(shí)間5%,提高產(chǎn)量5.6%。關(guān)鍵詞:功能基因亞功能片段粒子群優(yōu)化算法神經(jīng)網(wǎng)絡(luò)預(yù)測控制優(yōu)化Ab毗ractAbstractTheprojectisbasedonaChinese”863”projectn咖ed“Thepreparationofthesub一如nction‘segmentoffunctiongeneaJldthetechnologyplatfomlofqualitycontml”.Toensuretha
4、tthesystemgethighyield,thefernlentationprocessmustbeoptimized;fementationyjeIdandqualitymustbe.heightened.sothestrongemphasisofthisprojectislaidontheresearchofapplyingimprovedparticleswannoptimizationtotheoptimalcontrolofmaterialmakeupVelocity瓦,pHandDDof
5、enzymefermentationprocesscontr01.Comparingpanicleswannoptimizationwithotheroptimizationalgorithms,itsconvergencevelocityisfasterandoperationissimpIer.AccordingtotheproblemthatPS0algorit}1r11iseasytoplungeintothelocalminimum’sbasinwhenitdealswithmultidime
6、nsionalandcomplicatedproblems,aVariation—operatorPs0algorithm(VPSO)isdesigned.Thetestresultsoffourmultidimensionaltestmnctions,haVepmVedthatVPS0algorithmcanimproveme910balconvergenceVeIocityaJld910balsearchability,greatlye11}1ancetherateofglobalconvergen
7、ceandovercomettleshortcomingofbaSicPS0algoritllIll.ThedesignofoptimizationprojectisbasedonpredictivecontrolofBPneuralnetworkandpar鋤eteroptimizationofVPS0algoritllIll.AccordingtotheshortcomingthatconvergencevelocityofbasicBPalgorithmistooslow,PSOalgorithm
8、isappliedtooptimizeweightsofBPneuralnetwork.ThesimulationresultspmVethatnotonlyconve唱encevelocitybutalsoglobalsearchabi“tyofthealgorithmhavegoodeH色ct.Inordertovalidatethef音asibilityandvaIidityof、thisoptimizationproject,iti